When and how social movements mobilize action within and across nations to promote solidarity with refugees

When and how social movements mobilize action within and across nations to promote

solidarity with refugees

Thomas, Emma F.; Smith, Laura G. E.; McGarty, Craig; Reese, Gerhard; Kende, Anna; Bliuc,

Ana-Maria; Curtin, Nicola; Spears, Russell

Published in:

European Journal of Social Psychology

DOI:

10.1002/ejsp.2380

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2019

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Thomas, E. F., Smith, L. G. E., McGarty, C., Reese, G., Kende, A., Bliuc, A-M., Curtin, N., & Spears, R. (2019). When and how social movements mobilize action within and across nations to promote solidarity with refugees. European Journal of Social Psychology, 49(2), 213-229. https://doi.org/10.1002/ejsp.2380

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

R E S E A R C H A R T I C L E

When and how social movements mobilize action within and

across nations to promote solidarity with refugees

Emma F. Thomas* , Laura G.E. Smith†, Craig McGarty‡, Gerhard Reese§, Anna Kende¶ ,

Ana-Maria Bliuc‡, Nicola Curtin** & Russell Spears††

* School of Psychology, Flinders University, Adelaide, SA, Australia † Department of Psychology, University of Bath, Bath, UK

‡ School of Social Sciences and Psychology, Western Sydney University, Sydney, NSW, Australia § Faculty of Psychology, University of Koblenz-Landau, Koblenz, Germany

¶ Department of Social and Educational Psychology, Eotvos Lorand University, Budapest, Hungary ** Department of Psychology, Clark University, Worcester, Massachusetts, USA

†† Department of Psychology, University of Groningen, Groningen, the Netherlands

Correspondence

Emma F. Thomas, School of Psychology, Flinders University, Adelaide, 5042, SA, Australia.

E-mail: Emma.Thomas@flinders.edu.au. Received: 27 November 2016

Accepted: 6 February 2018 https://doi.org/10.1002/ejsp.2380 Keywords: social movements, social identity, emotion, ideology, solidarity, collective action, culture

Abstract

When and how do social movements form to mobilize action across national boundaries? In the context of the 2015 movement to support Syr-ian refugees, we develop an integrative model of transnational social move-ment formation shaped by pre-existing world-views (social dominance orientation and right-wing authoritarianism) and social media exposure to iconic events, resulting in an emergent group consciousness (“we are”, “we believe”, “we feel”). Group consciousness is, in turn, the proximal predictor of solidarity with refugees. Participants were from six countries: Hungary (N = 267), Romania (N = 163), Germany (N = 190), the United Kingdom (N = 159), the United States (N = 244) and Australia (N = 344). Multi-group structural equation models confirmed that Multi-group consciousness, shaped by individual differences and exposure to events through social media, was the proximal predictor of solidarity. The subjective meaning of group consciousness varied across samples, reflecting national differences. Results support the importance of considering individual and national dif-ferences, and group processes in understanding emergent social move-ments.

Social psychological models of collective action in one sense offer compelling explanations for collec-tive action and protest, but in another sense explain little about the actual emergence of action. — (Livingstone, 2014, pp. 124–125)

Over the last five years, the world has witnessed waves of social change movements that have crossed

national boundaries, such as the Arab Spring

(McGarty, Thomas, Lala, Smith, & Bliuc, 2014), the movement to arrest Ugandan war criminal Joseph Kony, 2012 (Kony 2014; Thomas et al., 2015), and the Occupy movement (Smith, Gavin, & Sharp, 2015). This article focuses on a recent, widespread, transna-tional social movement. In 2015, following years of extreme suffering by refugees, especially those escap-ing conflicts in Syria and Iraq, there was a widespread increase in efforts by members of the global commu-nity to address the plight of refugees, especially those seeking asylum in Europe (Timson, 2015). The increased support for refugees appeared to occur

immediately after the widespread social media dissem-ination of images of the body of a Syrian child, Aylan Kurdi, washed ashore on a Turkish beach on Septem-ber 15, 2015 (Vis & Goriunova, 2015). The image was popularly shared through social media and is credited with playing an important role in the dramatic social and political changes that ensued (Slovic, V€astfj€all, Erlandsson, & Gregory, 2017).

Social psychological science is in a strong position to explain why people participate in such social move-ments. As van Zomeren, Postmes, and Spears (2008) showed, people will act together to support a cause they are committed to, when they perceive that some-thing is wrong, and that they believe can succeed. However, social psychology has been less successful in explaining when these movements will form and how they do so (Livingstone, 2014; above). When does a stance, trend or thread of opinion become a social movement that provides a basis for united action? Similarly, current empirical and theoretical approaches to collective action are not well equipped to explain international, global, cross-cultural expressions of

collective action (van Zomeren & Louis, 2017). How do individual differences in worldviews about how the world should be intersect with exposure to iconic events to generate a social movement spanning many nations?

The current article takes up these challenges. We adopt an integrative approach, informed by the frame-work provided by Duncan’s (2012) integrated model of personality and social psychological theories of collective action, to address these questions about the emergence of the global solidarity movement to support Syrian refugees. We report on the findings from six nations (Hungary, Romania, Germany, the UK, the US, Aus-tralia) sampled in the immediate aftermath of the response to the image of Aylan Kurdi.

An Integrative Model of Solidarity with Refugees

The current research builds upon the framework pro-vided by Duncan’s integrative model of collective action to develop a model of solidarity with refugees. Duncan’s (2012) integrated model brings together research into the role of personality and life experi-ences in precipitating involvement in social move-ments, with current theorizing in social psychology about why people engage in collective action. Specifi-cally, the integrated model posits that key individual differences and life experiences act as antecedents to the formation of a group consciousness that, in turn, is the proximal predictor of collective action. We adapt this framework to explore the rapid emergence of transnational solidarity with refugees witnessed in October–November 2015.

Figure 1 provides a conceptual overview of our hypotheses. It shows that we conceive of the solidarity

movement (“outcomes”) as both psychological, reflecting a sense in which members of advantaged nations “stand with” those in the disadvantaged group (in this case, refugees); and behavioral, reflecting concerted actions to change the circumstances of a disadvantaged group (by taking solidarity-based collective action; Saab, Tausch, Spears, & Cheung, 2015; Thomas, Mavor, & McGarty, 2012). When people discuss solidarity they often invoke spatial metaphors: you “stand with” another, or you are “in solidarity” (e.g., Meadows, 2007), and this sense of shared fate with the disadvantaged group is key to current models of political solidarity (Reicher,

Cassidy, Wolpert, Hopkins & Levine, 2006; Subasic,

Reynolds, & Turner, 2008). Moreover, the key defining feature of this movement was that the popular support shown for the refugees through social media was instrumental in promoting the political changes that ensued (see Slovic et al., 2017; Vis & Goriunova, 2015). Accordingly, we conceive of solidarity also as including a set of behaviors designed to advance the circum-stances of a disadvantaged group (i.e., donating, advo-cating, volunteering; see Kende, Lantos, Belinszky, Csaba, & Lukacs, 2017; Saab et al., 2015; Thomas, McGarty, Rees, Berndsen, & Bliuc, 2016).

Where did this global solidarity movement come from? Figure 1 shows that we propose that exposure to the image of Aylan Kurdi and the plight of Syrian refugees acted as a signal life experience that, along-side individual differences in legitimizing ideology, promoted solidarity with refugees through the forma-tion of group consciousness.

Antecedents to Group Formation

What are the individual differences most likely to be implicated in longstanding support for (or opposition to) refugees? Here we focus on individual differences

g f e d c b a Antecedents to group formation: individual differences and life experiences

Psychological group formation: group processes Social movement formation: outcomes Individual differences Legitimizing worldviews (right-wing authoritarianism and social

dominance orientation) Psychological solidarity

(“we” stand with refugees) and behavioral solidarity (past, future and observed action to support refugees) Group consciousness

(social identification, group efficacy and moral group

emotions) Life experiences

Exposure to iconic events through global and social

media

Culture Shared systems

of meaning

Fig. 1: Conceptual adaptation of Duncan’s (2012) integrated model. Key individual differences relating to legitimizing worldviews (a; RWA and SDO) and exposure to iconic events through social media (b) precipitate an emergent group consciousness. Group consciousness (“we are”, “we believe” “we feel”) is, in turn, the proximal predictor of solidarity with members of disadvantaged groups (c). Culture (shared systems of mean-ing) qualifies the antecedents of group formation (a, b), the outcomes of group formation (g), as well as the nature of the group itself (d).

in legitimizing ideologies such as right-wing authoritari-anism (RWA; Altemeyer, 1998) or social dominance orientation (SDO; Pratto, Sidanius, Stallworth, & Malle, 1994) (see Figure 1, path a). We expect that higher levels of RWA and SDO will work against a pro-refugee group consciousness (see Stewart et al., 2016; Thomas, Cary, Smith, Spears & McGarty, in press). People high in SDO are unlikely to take action because they seek to protect their privileged status; while people high in RWA are unlikely to take action because asylum seekers are perceived to represent a threat to the social order (e.g., Cameron & Nickerson, 2009; Duckitt & Sibley, 2010; Reese, Proch, & Cohrs, 2014). While SDO and RWA are often highly corre-lated, the Dual-Process Motivational Model (Duckitt, 2001) suggests that these are unique, differential pre-dictors of prejudice against lower status or minority groups.

In relation to life experiences, Duncan (2012) articu-lated a role for broad experiences relating to family background, education and personal experiences. However, watershed moments can also provide the impetus for rapid mobilization (Duncan & Stewart, 2007; Minkoff, 1997). To the extent that iconic events can now be disseminated rapidly through global and social media, these foster what we term signal life expe-riences. Social media events can affect people in similar ways to other forms of direct experience and set in train similar processes to those enacted in off-line set-tings (e.g., Kende, van Zomeren, Ujhelyi, & Lantos, 2016; Thomas et al., 2015; Thomas et al., in press). A key difference is that these processes can happen more swiftly and be shared more broadly than was previ-ously possible (Castells, 2012). Social media provide spaces through which people can communicate a sense of grievance about the current status quo, and generate new identities (based on their opinions about how the world should be) through such social interac-tion (Smith, Thomas, & McGarty, 2015). The widely disseminated image of Aylan Kurdi created a wide-spread sense of injustice and grievance. The watershed of pro-refugee opinion that ensued in the online dis-cussion provided social validation of views (Turner, 1991) and became a vehicle for group formation (Kende et al., 2016; McGarty et al., 2014; Figure 1 path b).

In sum, key individual differences about what is right and just in the world (path a; SDO and RWA) and signal life experiences (path b; exposure to iconic images of the refugee crisis through social media) are expected to predict psychological group formation. Duncan (2012) broadly conceives of the individual dif-ferences and life experiences variables as independent predictors of group formation (as in Figure 1). How-ever, it is also possible that the two interact to explain those who identify as a supporters (and develop a group consciousness) versus those who do not. Here, several outcomes are possible based on the literature. One possibility is that, for people who had entrenched worldviews that opposed the movement of people

(high SDO and RWA), exposure to the issue through social media crystallized their non-support (a “polariz-ing” effect). Such a pattern would help to explain the widespread political polarization that the image pro-voked (Reese, Rosenmann, & McGarty, 2015). On the other hand, there is evidence that the image of Aylan Kurdi effectively mobilized large swathes of the popu-lation who had not previously been active on this issue (Smith, McGarty, & Thomas, 2018). It is there-fore possible that the image effectively “converted” those traditionally hostile to refugees, in a similar way that intergroup contact can have especially strong prejudice reduction effects on those high in RWA (but not SDO; Asbrock, Christ, Duckitt, & Sibley, 2012). In order to conduct a comprehensive test of the antece-dents of group formation, we tested whether legitimiz-ing ideologies and exposure to the image of Aylan Kurdi directly and interactively predict those who came to identify as supporters of refugees (vs. those who did not), and a pro-refugee group consciousness. An Emergent Group Consciousness

Although personality and life experiences may predis-pose a person towards taking action (or not), current theorizing within social psychology suggests that peo-ple will be more likely to become part of a social movement when they see themselves as united with others who share the same goal (Klandermans, 1997). Indeed, current approaches to collective action within social psychology emphasize that supporters must share a collective orientation to righting wrongs, and believe that they have the resources to act effectively (van Zomeren et al., 2008). Duncan (2012, p. 781; Duncan, 1999) terms this constellation of factors group consciousness, encompassing “social psychological vari-ables related to group identification and common fate, critical analysis of a group’s position in society, and a collective orientation toward redressing power imbal-ances between groups” (see also Bliuc et al., 2015; Fat-tori, Pozzi, Marzana, & Mannarini, 2015; Thomas, McGarty, & Mavor, 2016; Thomas, McGarty, Rees, et al., 2016). Figure 1 shows that group consciousness is conceptualized as a combination of social identifica-tion, moral emotions that capture a critical analysis of the group’s position (guilt, sympathy and outrage) and group efficacy beliefs that reflect the group’s beliefs about acting to address power imbalances (Duncan, 2012).

What is the nature of the social identity that under-pins the emergent movement? We propose that it is useful to consider social movements as groups based

on shared opinions (opinion-based groups; Bliuc,

McGarty, Reynolds, & Muntele, 2007). The opinion-based group concept is especially helpful where most supporters do not already belong to activist organiza-tions or instituorganiza-tions; in circumstances where support-ers are not acting in the interests of a social category that they belong to; and in cases where the longstand-ing experience of own-group adversity are not met

(see Simon & Klandermans, 2001; also McGarty, Bliuc, Thomas, & Bongiorno, 2009). These are the circum-stances that allow supporters to “become the change they want to see in the world” (Smith, Thomas, & McGarty, 2015).

People must also believe that their group’s actions can be effective in order to undertake coordinated action (van Zomeren et al., 2008) and beliefs about the ability of the group to respond effectively are a key component of group consciousness (Duncan, 2012). Whereas problem-focused coping (group efficacy) focuses attention on the instrumental strategies that are likely to improve the situation, emotion-focused coping regulates the group emotions tied to percep-tions of injustice (van Zomeren, Spears, Fischer, & Leach, 2004). Collective action research has typically focused on feelings of anger (e.g. Leach, Iyer, & Peder-sen, 2006; Mackie, Devos, & Smith, 2000) or moral outrage (Thomas & McGarty, 2009) as drivers of action, although other emotions are likely to be impli-cated in the regulation of disadvantage (van Zomeren et al., 2004) and can capture a critical analysis of the group’s position (Duncan, 2012). We therefore con-sider the role of three moral emotions as reactions to the injustice experienced by Syrian refugees: guilt, sympathy and moral outrage (Montada & Schneider, 1989).

Group-based guilt is an ingroup-focused emotion that arises from the appraisal of harm at least partly caused by the group’s actions (Leach, Snider, & Iyer, 2002), promoting behaviors that repair the wrongdoing rather than more wide-ranging forms of social justice action (Iyer, Leach, & Pedersen, 2004; see Thomas, McGarty, & Mavor, 2009b; for a review). Sympathy, on the other-hand, is outgroup-focused in that it focuses attention on the plight of the disadvantaged (Leach et al., 2002). Unlike guilt, sympathy does not attribute responsibility to any specific group or agent (Thomas et al., 2009b). Saab et al. (2015; also Thomas & McGarty, 2018) found that sympathy was a predictor of the tendency to support members of a disadvan-taged group. Finally, moral outrage is also other-focused but instead of focusing attention on the suffer-ing of the disadvantaged, it directs attention against a third party or authority held responsible for the disad-vantage (Leach et al., 2002; Thomas, McGarty, & Mavor, 2009).

The group consciousness concept therefore effec-tively captures (opinion-based) social identification, group efficacy and moral emotions (sympathy, guilt, outrage) as indicators of a latent factor (Figure 1). The relationships between these variables are known to be

bi-directional (see Thomas et al., 2012; Thomas, McGarty, & Mavor, 2009a for a review) and synthesiz-ing the variables in this way allows us to test more complex models than would be otherwise possible. We expect that group consciousness will be the proxi-mal predictor of psychological and behavioral solidar-ity with refugees (path c, Figure 1).

Social Mobilization Within and Across Nations The solidarity movement to support refugees in late 2015 was a global, transnational one, affecting social and political changes in the UK, Canada, the US, Aus-tralia, Eastern Europe, Western Europe, and beyond. The tensions between local and universal meanings are issues that have been widely considered in cross-cultural research on personality; however, they have not been systematically addressed in collective action research (see Stewart et al., 2016 for an exception). This omission is important because culture (shared sys-tems of meaning) fundamentally affects the emer-gence, and manifestation of, collective action (van Zomeren & Louis, 2017; captured in paths e, f, and g respectively, Figure 1). Is the mobilization process anticipated in Figure 1 universal, or are there local (nation-specific) differences? Answering this question raises difficult theoretical and empirical questions (van Zomeren & Louis, 2017).

Theoretically, one advantage of employing Duncan’s integrated model is that Duncan (2012, p. 784) emphasizes the phenomenological aspects of collective action. The group consciousness variable therefore empirically captures the key drivers of collective action (at the group level; i.e., van Zomeren et al., 2008), alongside a theoretical emphasis on the lived experi-ence of group membership too (see Drury, Cocking, Beale, Hanson, & Rapley, 2005). We would expect the phenomenological experience of the group to vary not least because of the diverging (social, political geo-poli-tical) circumstances of different nations reacting to the same events (path d, Figure 1). Emotions, in particu-lar, capture the lived experience of politics in people’s lives (Leach, 2010). They are markers of the legitimacy and stability of intergroup relations (see Leach et al., 2002) so it follows that emotions will vary where aspects of the intergroup context vary (see Mackie, Smith, & Ray, 2008, for a review). As such, the feel-ings of guilt, sympathy and outrage may act as mark-ers of qualitative differences in group consciousness across nations.

Similarly, it may be that some of the structural paths identified in Figure 1 play a relatively stronger or weaker role in some countries than others, speaking to important cross-cultural variations in the overall mobi-lization process. There has been shown to be cross-cul-tural variation in the relations between RWA, SDO and political attitudes respectively (Duriez, Van Hiel, & Kossowska, 2005, Figure 1, path f); as well as in the relationship between social media use and the

1_{Opinion-based identities can be politicized identities, in the sense}

that they incorporate an awareness of the power relations in which a struggle is embedded (Simon & Klandermans, 2001); however, in this case, the term “politicized identity” does not sufficiently differentiate supporters of refugees from opponents of refugees, both of whom can both be acting in terms of a politicized identity but advocate funda-mentally different visions of social relations.

organization of social movements (Harlow & Harp,

2012; Figure 1, path e). Fischer, Becker, Kito, and

Zamantılı Nayır (2017) observed that cultural struc-tures relating to self-construal (interdependent vs. independent; Markus & Kitayama, 1991) and face concern shape the form of collective action taken against sexism across different nations (path g).

Empirically, questions about differences in phe-nomenology and subjective meaning are typically addressed through qualitative forms of analysis (e.g., Drury et al., 2005). We take a different approach and tackle these questions of cross-cultural similarity and difference using tests of measurement invariance (Steenkamp & Baumgartner, 1998). It is recognized that a given construct is stable across culture, and mean-level comparisons (e.g., of group consciousness) should only be made where the loadings on that latent factor are the same across samples (metric invariant) and the intercepts are the same (scalar invariant; Dim-itrov, 2010). By contrast, significant differences in the loadings of a latent factor suggest qualitative differ-ences in the meaning attributed to that construct (Vandenberg & Lance, 2000). Although establishing measurement invariance is usually viewed as pre-con-dition for cross-cultural or longitudinal comparisons, we suggest that it might be useful in its own right to map out similarities and differences in the content of a group consciousness latent factor. However, one ten-sion here is that partial metric invariance is a pre-con-dition for multi-group structural equation models (Vandenberg & Lance, 2000). That is, a basic level of (configural and partial metric) invariance must hold in order for cross-national comparisons in model paths to be meaningful (see Davidov, D€ulmer, Schl€uter, Sch-midt, & Meuleman, 2012). The implication is that our statistical methods can accommodate some cross-cul-tural difference but not absolute differences.

The Current Research

In September–October 2015, we sampled six countries

that varied in their proximity to the movement of refugees and political responses to the refugee crisis: Hungary, Romania, Germany, the UK, the US and Australia. Participants were recruited in the aftermath of the response to the picture of Aylan Kurdi, but prior to the November 13 Paris terrorist attacks. In line with our conceptualization of solidarity as both psychologi-cal and behavioral (Figure 1), we adapted a pictorial measure of self–other overlap (Schubert & Otten, 2002) as a measure of the degree to which members of relatively advantaged groups (i.e., nations) experience psychological solidarity with refugees. We also took measures of the degree to which participants had already engaged in behaviors that reflect efforts to improve the circumstances of refugees, the degree to which they intended to in the future, and a measure of observed behavioral support for refugees. We test two sets of hypotheses based on Figure 1. Our research

is exploratory in that, in most instances, we do not specify formal a priori hypotheses about the specific pattern of differences across the six samples. Rather, our focus here is on demonstrating the utility of this framework for theorizing and empirically testing cross-cultural differences in the formation and outcomes of social movements.

The first set of hypotheses relate to the antecedents to group formation. Key individual differences about what is right and just in the world (legitimizing ideolo-gies; SDO and RWA, path a) and signal life experi-ences (exposure to iconic images of the refugee crisis through social media, path b) were expected to predict refugee group consciousness. This emergent, pro-refugee group consciousness is then the proximal pre-dictor of solidarity (see Figure 1, path c).

We tested the antecedents of group formation in two ways. The first was to look at the role of legitimiz-ing ideologies and signal life experiences, and their product (interaction term), in differentiating those who identify as supporters of refugees (vs. those who did not), controlling for culture. Our analysis tests the competing possibilities regarding the direct and inter-active roles of legitimizing ideology and exposure to the iconic image: specifically, whether ideology and exposure do indeed combine to polarize people based along pre-existing positions, or convert those espe-cially hostile to outgroups. In a second substantive test of the antecedents of group formation, we modeled legitimizing ideologies and exposure to the iconic image as predictors of a latent group consciousness fac-tor using a sub-sample of self-identified supporters. Doing so allowed us to test whether the antecedents to group formation were the same across national con-text, or whether there was cross-cultural variation in that process (paths e and f). We did not make specific a priori predictions about whether those paths would be the same or different across the six samples. How-ever, one possibility is that the dissemination of the image through social media played a less important role in Hungary where refugees entered as a point of transition and were therefore highly visible in the off-line environment (cf. Harlow & Harp, 2012).

Our second set of hypotheses related to the cross-cultural conceptualization of group consciousness and its relation with solidarity. We expected that the same key components of group consciousness (identifica-tion, group efficacy, moral emotions of sympathy, guilt and outrage) would load significantly across nations (demonstrating configural invariance). Moreover, in keeping with the tenets of the social identity approach (Turner, 1991), we predicted that social identification would also be stable (metric invariant) across context, reflecting its central role in motivating collective action (van Zomeren et al., 2008). Thus, the subjective unity and common fate of the movement across great national diversity stems from shared (invariant, uni-versal) group membership as supporters of refugees.

However, in keeping with the ideas articulated above about moral group emotions as markers of

intergroup relations, we expect that they (guilt, sym-pathy, and outrage) and group efficacy would be vari-ant, yielding differences in the meaning of group consciousness (metric non-invariant). There were also expected to be mean level differences (scalar non-invariant) in emotions and efficacy across national samples. Thus, we did not expect to observe full mea-surement invariance because the measures are them-selves sensitive to prevailing social, historical and (geo) political realities. It may be that, for instance, group consciousness is flavored more strongly by feelings of guilt in countries that had previously held retributive political stances towards refugees (e.g., Australia, Romania, Hungary, and the UK) than in other sam-ples, capturing illegitimate harm for which one’s own group is (at least partially) responsible (Thomas et al., 2009b). The need for pragmatism may also be stronger or weaker depending on the national context (e.g., how responsive the political authorities are), and this would be reflected in differences in the loading of effi-cacy across nations. In this way, we theoretically and empirically account for a sense of unity across the national borders (generated by a universal or invariant social identification indicator) but also map out poten-tial differences captured by diverging strength of the moral emotions and/or efficacy.

Having identified which of the aspects of the group consciousness construct were the same/different across national contexts, we then test the full model antici-pated in Figure 1 using multi-group structural

equa-tion models. Cross-national differences in the

manifestation of solidarity are reflected in stronger/ weaker paths from group consciousness to solidarity on some markers compared to others (path g). Con-versely, it was plausible that the model would be

gen-eralizable across samples indicating that group

consciousness predicted different forms of solidarity relatively universally.

Method Participants

Data were collected in September–October 2015.2

Par-ticipants were 1,367 people from six countries:

Hun-gary (N= 267), Romania (N = 163), Germany

(N= 190), the United Kingdom (N = 159), the United

States (N= 244) and Australia (N = 344). We

excluded participants who had missing responses for

over 50% of the questions (n= 136; primarily people

who had clicked on the link but not commenced the

survey) or who completed the survey in <4 min

(n= 4) – the sample sizes reported above are the final

samples. These samples provide sufficient power to

detect small effects (.10), with power= .80 and

a = .05, in models with one latent variable and eleven observed variables (Soper, 2016). Participants were primarily female (59.47%; 30 people did not report their gender) and the average age was 30.77 (SD =15.2). Table 1 displays the demographic information for each of the samples.

Procedure and Measures

The questionnaire was originally constructed in Eng-lish and the key items were altered to reflect the rele-vant country. For the German, Hungarian and Romanian samples the survey was back-translated by at least two independent translators; disputes were resolved through discussion. In Australia, the US, Romania and Hungary, data were collected online, while in Germany the data were collected through pencil and paper questionnaires; the UK participants chose to participate either online or via hard copy. In Australia, Romania, the US and Germany, the sample comprised both members of the general public and students; the Hungarian and UK sample included only students who participated for course credit (Hungary) or as part of routine class activity (UK; see Table 1). In Australia, Romania and the US, members of the gen-eral population were accessed by circulating a link to the questionnaire (titled “Attitudes about the Syrian refugee crisis”) through personal networks and/or posting on the webpages of public forums (e.g., Craigs-list). In Germany, participants were approached on campus and in public places of a university town such as the train station and asked if they wished to partici-pate. Participants responded on a 7-point Likert type

scale (1= Strongly disagree, 7 = Strongly agree) unless

otherwise indicated below. Please contact the lead author for copies of the questionnaire (and transla-tions) and/or raw anonymized data.

Measures of Individual Differences and (signal) Life Experience

Image and (social) media exposure. We consid-ered the role of (social) media in two ways. To avoid experimental demands that could arise from referring directly to the image of Aylan Kurdi, we asked partici-pants whether they had “seen a recent image (photo) of the refugee crisis that has powerfully affected your views about this issue”; participants responded yes (=2) or no (=1).

Second, to assess exposure to the issue through (so-cial) media more generally we asked two questions to assess both the breadth and depth of exposure. To assess breadth (in terms of the range of media) we asked: “What is the main way that you have learned about the recent events of the Syrian refugee crisis?” Participants selected all that applied from television, newspaper, social media, radio, talking to other

2

Specifically, in Australia the data were collected between Septem-ber 16 and OctoSeptem-ber 5; in the United Kingdom between OctoSeptem-ber 1 and October 8; in the United States between September 22 and October 8; in Romania between October 15 and October 26; in Hungary on September 21; and in Germany between September 29 and October 25.

people, and other media. Participants were allocated a point for each media selected, summed such that the more media selected, the greater the breadth of expo-sure (values ranged from 0 to 6). To assess depth of exposure we also asked about the time spent engaging with this issue on social media: how much time they had spent in the past week reading online discussions

or watching videos about the refugee crisis

(none= code 1, 0–15 min = 2, 15–30 min= 3,

30 min-1 hr= 4, 1–3 hrs = 5, 3+ hrs = 6). We

aggre-gated responses to these two questions to form a sec-ond measure of global and social media exposure.

Legitimizing ideology. We used a single item from the social dominance orientation scale (SDO; Pratto et al., 1994) and a single item from the right

wing authoritarianism scale (RWA; Altemeyer,

1998). The items were: “It’s OK if some groups have more of a chance in life than others” (SDO) and “The only way our country can get through the cri-sis ahead is to get back to our traditional values, put some tough leaders in power, and silence the trou-blemakers spreading bad ideas.” The items were

moderately correlated, r= .32, p < .001. Since they

reflect different motivational states (Duckitt & Sibley, 2010), we retained them as separate predictors in our tests of the model.

Measures of Group Consciousness

Emotions: Guilt, sympathy and outrage. Two items measured the emotions of guilt, sympathy and outrage, respectively. Participants read “Considering the plight of Syrian refugees, I feel: Sympathetic [com-passionate]; guilty [responsible]; outraged [angry].” The internal validity for all three scales was acceptable in each of the samples (a = .78 to .95 sympathy; a = .60 to .89 guilt; a = .84 to .92 outrage).

Group efficacy. Two items measured group effi-cacy: “Together [national group members] can make a positive difference for Syrian refugees” and “Together [national group members] can improve the outcomes

for Syrian refugees”, a = .88 to .98 for each of the

samples.

Social identification. Participants were first asked to respond to a categorical item that assessed self-cate-gorization as a supporter of Syrian refugees. “Please think about how you see yourself in relation to Syrian refugees. Do you identify as a supporter of global

action to support Syrian refugees?” Participants

responded “yes” or “no”. Participants who indicated that they did not identify as a supporter did not com-plete further measures of pro-refugee social identifica-tion. However, if they selected “yes” to the categorical question, participants were then asked to complete further measures of the degree of identification. Three items (adapted from Cameron, 2004) measured identi-fication as a supporter of Syrian refugees (only for sup-porters): “I see myself as a supporter of Syrian refugees”; “I identify with other supporters of Syrian refugees”; “Supporters of Syrian refugees have a lot to

be proud of”,a = .68 to .82 for each of the samples.

Measures of Psychological and Behavioral Solidarity

Psychological solidarity. Participants were pre-sented with seven pictures of two increasingly overlap-ping circles; one circle was labeled [nation] and the other was labeled Syrian refugees (Schubert & Otten, 2002). Participants read the information that “The cir-cles represent different levels of closeness between the two groups” and were asked to select the picture that best represents how close they, as a member of their national group, feel to Syrian refugees.

Table 1. Demographic information and descriptive statistics (mean, standard deviations) for the key variables for the six samples Hungary N = 267 Romania N = 163 Germany N = 190 United Kingdom N = 159 United States N = 244 Australia N = 344 % Female 78.7 76.7 51.1 82.4 69.7 23.3 Age 21.22 (2.48) 37.4 (10.80) 27.51 (9.55) 18.5 (1.48) 40.08 (21.97) 36.5 (15.07) % Students 100 100 36.3 96.9 42.2 14.2 % Supporters 37.5 44.2 81.1 90.6 82.4 82.6 Social dominance 2.89 (1.61) 3.62 (1.77) 2.63 (1.50) 2.53 (1.49) 2.65 (1.72) 2.64 (1.79) Right-wing authoritarianism 3.26 (1.75) 3.70 (1.93) 2.81 (1.62) 2.32 (1.34) 2.65 (1.81) 2.88 (2.09)

% People who have seen image 64.4 54.6 29.5 68.6 72.5 78.2

Overall media exposure 3.06 (0.76) 2.21 (0.92) 2.89 (1.06) 2.01 (0.76) 2.29 (0.93) 3.09 (0.97)

Social identification 5.61 (0.75) 5.04 (0.98) 4.80 (0.96) 5.47 (0.87) 5.40 (0.99) 6.03 (1.00)

Group efficacy 3.42 (1.38) 3.82 (1.68) 4.85 (1.50) 5.86 (1.09) 5.44 (1.30) 5.71 (1.66)

Guilt 3.05 (1.37) 2.86 (1.34) 3.78 (1.31) 3.70 (1.48) 3.45 (1.71) 3.77 (1.97)

Sympathy 4.54 (1.29) 5.36 (1.37) 5.69 (1.07) 6.27 (0.81) 5.95 (1.21) 6.05 (1.64)

Outrage 3.38 (1.56) 3.80 (1.67) 4.21 (1.69) 5.17 (1.32) 5.17 (1.61) 5.45 (1.77)

Psychological (self-other) solidarity 2.61 (1.35) 2.54 (1.45) 3.36 (1.33) 2.94 (1.50) 3.06 (1.70) 3.61 (2.07) Past (self-reported) solidarity 1.95 (0.13) 1.94 (0.14) 1.83 (0.20) 1.82 (0.23) 1.82 (0.26) 1.56 (0.32)

Future (intended) solidarity 3.26 (1.35) 3.24 (1.43) 3.78 (1.12) 4.63 (1.17) 4.37 (1.36) 4.98 (1.75)

Behavioral solidarity: Past, future and observed behavior. We measured behavioral soli-darity as past (self-reported) action, future action (ac-tion inten(ac-tion) and observed ac(ac-tion. Participants responded to four items: “I intend to sign a petition”

[and self-report: “I have already. . . etc.”]; “I intend to

support the plight of Syrian refugees by donating money to the cause”; “I intend to post on social media about this issue”; “I intend to volunteer to support

refugees” on a 1–7 Likert-type scale indicating their

agreement with the intention items, and indicated yes

(=2) or no (=1) for the self-report items.3Scales were

internally consistent for future intended action, a = .62 to .84, for past self-reported action in Australia

and the US, botha = .62. However, past self-reported

actions were not internally consistent in the other samples (a = .21 to .47), probably because dichoto-mous data reduce variability (inter-item correlations). Past self-reported action was significantly correlated

with both the intended behavior (r= .59, p < .001)

and observed behavior (see below; r= .15, p < .001),

providing convergent validity for the scales. Since peo-ple with higher scores report performing more of the relevant behaviors, we therefore created the scales as an index of past commitment.

These measures of past and future action were sup-plemented with a measure of observed behavior. Specifically, participants were told that the researchers would make a donation of US$1 on their behalf. They were then asked to allocate how much of that US$1 they would like to be sent to an accredited agency working to assist Syrian refugees, with any remainder to be allocated to an agency working to assist disad-vantaged children within [the home nation]. This donation allocation thus constitutes a measure of observed behavioral support for refugees relative to another worthy cause. Researchers made these

pay-ments on behalf of participants.4

Results Analytical Strategy

Our analysis comprised three parts. First, we used the entire sample to focus on distinguishing people who saw themselves as supporters of action on refugees versus those who did not. We expected that key indi-vidual differences (legitimizing ideology, path a) and media exposure (both to the specific image and

over-all, path b) would predict psychological group

formation both uniquely and, prospectively, in inter-action. Therefore, we conducted logistic regression analyses (controlling for non-independence) using SDO, RWA, and media exposure as independent vari-ables to predict whether or not participants self-cate-gorized as a supporter of Syrian refugees (see Figure 1).

In the second and third sections, we included only people who identified as supporters to test the full integrative model of solidarity across the six nations

(n= 955). This is because people who did not

self-categorize as a supporter of refugees did not complete measures assessing the degree of their social identifica-tion. The second section tests for the measurement equivalence of group consciousness across nations using tests of measurement invariance following pro-cedures outlined by Vandenberg and Lance (2000; measurement model, path d). In the third section, we used a multi-group structural equation model to test the full integrative model (paths a, b, c), and for national differences in the structural paths (structural model, paths e, f, and g).

In both sections 2 and 3 the multi-group analyses employed a similar strategy. We adopted the preferred forward-testing method (Vandenberg & Lance, 2000), first estimating a baseline model in which all paths, intercepts and latent means were allowed to vary freely, and subsequently testing increasingly con-strained models to see if fit significantly improved or deteriorated using tests of the chi-square difference. Invariance in measurement across countries is indi-cated if the more constrained model/s fit/s better than when indicators vary freely across country (section 2, path d). Similarly, evidence that there are different rel-ative weightings of the paths in the model (paths e, f, g) is suggested if the constrained model fit is better than the model where the paths are allowed to vary (section 3). Evidence of good fit is indicated by a

non-significant chi-square, a CFI≥ .95, a RMSEA ≤ .08,

and a SRMR≤ .08.

Preliminary Analyses

Missing values analysis revealed a small amount of

data (<5% for each variable in all the samples) that

were missing completely at random (MCAR; all ps: .11 to .63) except in the German sample, which was not

MCAR,v2(804)= 1001.89, p < .001. Since there was

negligible missing data from the German sample (<5%; see Enders, 2010) we nevertheless treated all values using Full Information Maximization Likeli-hood in MPlus version 7.2. Table 1 displays the means (standard deviations) for the key variables across the six samples.

The Antecedents of Psychological Group Formation

We first conducted multi-level binomial logistic regres-sion exploring the antecedents of psychological group

3_{In the Romanian sample there was a data collection anomaly such}

that participants responded to the self-report items on a 7-point Lik-ert-type scale (Strongly Disagree–Strongly Agree), instead of on a bin-ary (yes/no) scale as in the other samples.

4_{Payments were made as directed by participants to the International}

Red Cross, as well as the national division of Save the Children for all of the countries except for Hungary, for which a donation was made to the Polgar Foundation for Equal Opportunities.

formation to test whether these predicted self-categori-zation as a supporter (vs. not). For completeness, we also tested for interactions between ideology (SDO/ RWA) and media exposure (image/overall) using the procedures identified by Aiken, West, and Reno (1991). Independent variables were centered and interaction terms were created based on their multipli-cation. Since the data were nested at the nation-level, we controlled for non-independence of observations using the “complex” command in MPlus; doing so allows us to take into account the non-independence of observations using a sandwich estimator, but it does not specify a model for the between-nation level (Muthen & Muthen, 1998-2012).

In the first step, social media (image and overall exposure) and ideology (SDO and RWA) terms were entered as predictors of the dependent variable,

self-categorization (coded no= 1, 2 = yes). In the second

step, we entered the four interaction terms between ideology (RWA and SDO) and the two media vari-ables. We found support for the idea that legitimizing ideology was a significant negative predictor of

mem-bership, c = .26, SE = .06, p < .001 for SDO,

c = .25, SE = .03, p < .001 for RWA (path a). Expo-sure to iconic images marginally predicted

self-categor-ization as a supporter,c = .13, SE = .07, p = .057, but

overall media exposure did not, c = .07, SE = .09,

p= .42 (path b). However, as expected, there was a

significant two-way interaction between SDO and

media exposure, c = .09, SE = .03, p = .009; there

were no other significant interactions (all ps> .14).

Figure 2 displays the interaction. It shows that overall media exposure did not affect the likelihood of being a supporter for those low in SDO, unstandardized c = .06, SE = .14, p = .64. For those high in SDO, however, higher social media exposure was associated with a reduced likelihood of identifying as a supporter,

unstandardizedc = .18, SE = .08, p = .02.

Thus, ideologies that legitimize inequality (RWA and SDO), both directly and in combination with exposure to the issue through social media, signifi-cantly reduced the likelihood that participants

self-categorized as a supporter of Syrian refugees (paths a, b). We note that the image predictor was a marginal predictor but may have been weak because it is cate-gorical (yielding a catecate-gorical IV predicting a categori-cal DV) and had truncated range; Table 1 shows that, for all the samples except Germany, the majority of people reported having seen powerful images of the crisis.

Group Consciousness Across Nations: A Test of Measurement Invariance

Our initial test of measurement invariance (path d) specified a latent factor model with five indicators: identification, group efficacy, and the three emotions (guilt, sympathy and outrage). We first estimated the model separately for each of the six samples. Fit was poor for several of the samples and the combined

(baseline) chi-square was large and significant, v2

(30)= 84.276, p < .001. Outrage was a non-significant

predictor in the Hungarian and Romanian samples and a weak predictor in the UK sample. Since a similar pattern of zero and non-zero loadings across samples (i.e., configural invariance) must be met in order to conduct subsequent tests of measurement invariance and multi-group models (Steenkamp & Baumgartner, 1998), we removed outrage and focused our test of the emotional predictors of sympathy and guilt. Accordingly, our (revised) initial test specified a latent model with four indicators: identification, sympathy, guilt and efficacy. The results of the tests for each step of measurement equivalence are displayed in Table 2. The details of all the specific models that were tested (i.e., whether each of the indicators was different within each level of measurement invariance) can be found in the Supporting Information.

We first tested the model with unconstrained factor loadings and intercepts. Fit was acceptable to excellent

for all the samples,v2(2) = .82 to 7.05, and the

com-bined (baseline) chi-square was moderate and only

marginally significant, demonstrating configural

invariance (Model 1, Table 2). Thus, the model is valid for each of the national samples.

In order to test whether the weightings of indicators were the same or significantly different (metric invari-ant), we next compared this model with one in which the parameters for the indicators were constrained to be the same across the six samples (path d). In order for the model to be identified, we had to set one path to unity: we selected identification because it had the most similar unstandardized regression weights across nations (see Dimitrov, 2010). Overall model fit deteri-orated (Model 2a, Table 2). The difference between

Model 1 and 2a was significant, Δv2 (15)= 27.32,

p= .03, confirming that the meaning of the latent

variable was different for the different samples (metric non-invariance). The best fitting model (see Model 2b, Table 2) held sympathy and identification invariant across samples, with guilt and efficacy allowed to differ

across samples (see Supporting Information for

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Low Social Dominance Orientation

High Social Dominance Orientation Pr obability of classification as supporter Low Media Exposure High Media Exposure

Fig. 2: Probability of classification as a supporter of Syrian refugees decreases for those who are high in social dominance orientation and who report high levels of media exposure.

details). Thus, the factor loadings are identical (invari-ant) for sympathy and identification across the national samples, but differ substantively across nations for guilt and efficacy. This suggests that the meaning of the latent construct (group consciousness) is different across national contexts because it is indi-cated more strongly by guilt/efficacy in some places than others.

We next sought to establish whether the intercepts were the same across samples (scalar invariance). To do so we compared the best-fitting metric invariance model (Model 2b) with a model in which the inter-cepts were constrained to be the same. Overall model fit deteriorated dramatically (see Model 3, Table 2), and the difference between Model 2a and 3 was

signif-icant,Δv2(20)= 437.33, p < .001, indicating that the

levels of the items are different across the samples (sca-lar non-invariant). That is, the mean levels of the underlying items (guilt, efficacy, identification, sympa-thy) varied significantly across samples. The best-fit-ting scalar invariant model allowed all four items to be different (see Supporting Information) and was thus identical to Model 2b.

The final measurement model incorporated partial measurement invariance (for the factor loadings of sympathy and identification) but allowed the other factor loadings, intercepts and residual variances to vary freely (path d). Substantively, the model shows that identification and sympathy were stable across national context, but that guilt and efficacy favored the group consciousness more strongly in some nations than others (see Figure 3); there was varia-tion in the levels of each of the items across navaria-tions. This partial measurement invariance allows for tests of the structural model but precludes cross-cultural comparisons of mean levels of the latent variable group consciousness (Dimitrov, 2010).

Testing the Integrative Model Across Nations: Multi-group Analyses

Having identified a suitable measurement model for the group consciousness factor, we adopted a similar approach in our tests of the full structural model. We first tested a baseline model in which all the structural

paths were freely estimated. RWA, SDO, exposure to the image, and social media exposure predicted the group consciousness latent factor (that was indicated by identification, guilt, sympathy and efficacy). Group consciousness predicted psychological solidarity, as

well as past, future and observed solidarity action5

Fit was acceptable, v2 (258)= 424.40, p < .001,

RMSEA= .07, CFI = .93. SRMR = .06, though the

CFI indicated that there was scope for improvement. Inspection of the standardized pathways showed that RWA was not a unique predictor of group conscious-ness in any of the samples except for Hungary

(b = .25, p = .01). Given that it played a muted

inde-pendent role in these data, we removed RWA from our test of the integrative model.

Our test of the model (without RWA) again

demon-strated acceptable fit, v2 (216)= 360.67, p < .001,

RMSEA= .07 [90% CI: 0.05, 0.08], CFI = .94,

SRMR= .06. The chi-square statistic and indices based

on its distribution (the CFI) are sensitive to large sam-ple sizes, and the other indices suggested good fit (the RMSEA and the SRMR; Kline, 1998). This therefore became our baseline model. Figure 4 shows the pat-tern of standardized regression weights when they are not constrained to be the same for the six samples. Fig-ure 4 shows that exposFig-ure to the iconic image and social media are weak positive predictors of group

con-sciousness (path b), whilst SDO was a weak–moderate

negative predictor (path a). Group consciousness was strongly indicated by social identification, guilt and sympathy, and more weakly indicated by group effi-cacy. Group consciousness, in turn, was a moderate to strong predictor of psychological solidarity, past, observed and future solidarity action (path c).

Having established that the model fitted with the different samples, we next compared this baseline model with one in which the paths were constrained to be the same across the six samples as a test of cross-cultural difference (paths e, f, and g). Overall model fit

Table 2. Fit statistics for tests of measurement invariance for the group consciousness factor across the samples Interpretation of the different steps of

measurement invariance testing v2(df) p RMSEA CFI SRMR

Model 1 Configural invariance Same pattern of significant indicator to factor loadings across samples. The model is valid for each of the nations.

v2_{(12)= 20.79, p = .05 .07 .99 .03}

Model 2a Metric invariance Factor loadings are the same across samples. The respondents attribute the same meaning to the latent construct.

v2_{(27)= 48.11, p = .007 .07 .97 .11}

Model 2b Best fitting metric invariance model _v2_{(18)= 26.64, p = .09 .06 .99 .05}

Model 3 Scalar invariance The intercepts are the same across samples. The levels of the underlying items are equal across nations.

v2_{(38)= 463.96, p < .001 .27 .44 .37}

5

For completeness, we also tested a model that included the interac-tion terms between image and SDO, RWA, and overall media expo-sure and SDO, RWA, as predictors of group consciousness in each of the samples. None of the interaction terms were significant except for that between overall media exposure and SDO in Romania (b = .58, p< .001).

deteriorated, v2 (251)= 414.48, p< .001,

RMSEA= .07, CFI = .93, SRMR = .10. The difference

between the two models was significant, Δv2

(35)= 53.81, p = .02, confirming that there are

differ-ences in the relative weightings of the paths in the model across the different national samples. That is, there is evidence of cross-national variability in the model (supporting paths e, f, g, Figure 1).

In order to determine the nature of those differences we systematically released each of the pathways to see if fit improved significantly compared to the baseline model. There was no evidence of cross-national vari-ability for the paths from SDO to group consciousness

(path f, p> .05) and overall media exposure to group

consciousness (path e, p> .05), however fit improved

significantly for the path from image to group

con-sciousness, Δv2 (30)= 46.44, p = .03 (path e). This

suggests that SDO and social media exposure were related to group consciousness in a relatively universal fashion but that there were national differences for

the relationship between exposure to the image and group consciousness.

In terms of the relationship between group con-sciousness and solidarity outcomes (path g), fit also improved when the paths from group consciousness to

psychological solidarity, Δv2(30) = 48.16, p = .02, as

well as past, Δv2 (30) = 48.89, p = .02, future Δv2

(30)= 48.05, p = .02, and observed behavior Δv2

(30)= 52.66, p = .01 were released, indicating that

there is variation in all of these paths across the sam-ples. Our final tests constrained paths to be similar/dif-ferent based on the weightings observed in Figure 4. Holding similarly weighted paths to equality (but allowing them to differ across countries) did not signif-icantly improve fit above the models in which all the paths varied freely. Since including the additional con-straints did not improve fit, we retained the more par-simonious model for each pathway (allowing them to

vary freely), v2 (226)= 389.48, p< .001,

RMSEA.= 07 [90% CI: 0.06, 0.08], CFI = .93,

Observed Solidarity (Allocation) Image (1 = no, 2 = yes) Social Media Exposure Social Dominance Orientation Group Consciousness Psychological Solidarity (Self-reported) Past Solidarity Action Future Solidarity Action (Intention) Group Efficacy

Social Identification Guilt Sympathy

RO HU GER UK US AUS .516, .740, .657, .772, .758, .837

Individual Differences and Life Experiences Psychological Group

Formation Social Movement Formation RO HU GER UK US AUS .633, .505, .730, .644, .503, .569 RO HU GER UK US AUS .277, .633, .551, .592, .697, .743 RO HU GER UK US AUS .362, .299, .314, .554, .460, .687 RO HU GER UK US AUS .241, -.016, .206, .154, .253, .096 RO HU GER UK US AUS .388, .137, .190, .352, .379, .133 RO HU GER UK US AUS .003, -.373, -.365, -.295, -.134, -.387 RO HU GER UK US AUS .537, .695, .538, .436, .421, .462 RO HU GER UK US AUS .146, .417, .360, .352, .418, .506 RO HU GER UK US AUS .407, .319, .448, .532, .467, .433 RO HU GER UK US AUS .835, .680, .676, .798, .818, .747

Fig. 4: Weightings of the measurement and structural paths for the integrative model for the six samples, for self-identified supporters only (n_{= 955).} RO HU GER UK US AUS .878, .777, .618, .738, .818, .861 RO HU GER UK US AUS .520, .477, .683, .623, .411, .532 RO HU GER UK US AUS .419, .627, .546, .630, .802, .714 RO HU GER UK US AUS .317, .269, .462, .647, .447, .686 Group Consciousness Group Efficacy

Social Identification Guilt Sympathy

Fig. 3: Loadings of the four indicators (social identification, guilt, sympathy and group efficacy) on the group consciousness factor for the six samples when allowed to vary freely. Model includes self-identified supporters only (n_{= 955).}

SRMR= .07. Overall, Figure 4 provides good support for the key hypothesized paths in the integrative model; it also illustrates that there was both cross-cul-tural variability and stability in the mobilization pro-cess.

Discussion

When do social movements form in relation to iconic events? How do these processes transcend national, cultural and geographical boundaries to effect global social change? Our research considers behavioral and psychological solidarity with Syrian refugees as an out-come of individual differences in legitimizing ideology, exposure to iconic events through global and social media, and group processes (the formation of a group consciousness). Across six different national samples (with diverging geo-political obligations and responses to the crisis) obtained in the aftermath of the global reaction to the image of Aylan Kurdi, we showed that social movements form when exposure to iconic events through social media intersect with pre-existing world-views (especially SDO) to catalyze the forma-tion of new groups (paths a, b, Figure 1). These new groups were defined along opinion-based fault lines: people who support, or do not support, action to assist Syrian refugees (Bliuc et al., 2007).

We also sought to tackle one of the key challenges raised by these transnational phenomena: how can we understand those factors that promote a subjective sense of common cause, whilst at the same time acknowledging variation in those processes across national contexts? To that end, we adapted Duncan’s conceptualization of group consciousness as a latent factor indicated by social identification, moral group emotions and group efficacy beliefs. We adopted methods routinely taken up in cultural psychology to show that the same pattern of indicators universally (significantly) captured group consciousness (configu-ral invariance), and also that social identification and sympathy were two factors that were stable across the six samples considered here (they showed metric invariance). However, the importance of feelings of guilt and efficacy beliefs to the group consciousness factor varied systematically across the samples, pro-ducing a group consciousness that was different across nations (metric non-invariant). Thus, the meaning of the group consciousness was flavored more strongly by guilt in Germany and the UK; and by relatively greater efficacy in the UK and Australia (path d, Fig-ure 1).

Finally, we also sought to consider whether the pro-cesses of mobilization were the same or different across national contexts (van Zomeren & Louis, 2017). Here we again found evidence of cross-cultural stabil-ity and variabilstabil-ity in the importance of specific paths. Exposure to iconic images of the crisis through global and social media was more important in the formation of a group consciousness in Romania, the UK and the

US, and relatively less important for participants in Australia, Germany and Hungary (path e, Figure 1). Figure 4 shows that the strength of the relationship between media exposure and group consciousness, and SDO and group consciousness, respectively, varied from zero to moderate. However, the tests of the full model included only self-assigned supporters so the variability in the independent variables was likely to have been restricted relative to tests that included the full sample. Similarly, the strength of the relationship between group consciousness and solidarity outcomes (psychological and behavioral) was also different across national contexts (path g, Figure 1). Psychologi-cal solidarity was more strongly associated with group consciousness in Hungary; past self-reported action was more important in the US and Australian samples, whilst future intention was less central in the German and Hungarian samples (see Figure 4). Overall, we found strong support for our hypotheses about the ways in which individual differences, life experiences and group processes combine to explain how social movements form within and across nations.

Why Now? Discontinuity in Social Movement Formation

Livingstone (2014) considered whether social psycho-logical models can tell us something about the point at which protest will emerge, ultimately concluding that existing models tell us little about the tipping points in collective action (see also McGarty et al., 2014). The current research conceives of rapid social movement formation as a process of novel group formation, argu-ing that in order to understand the generation of new phenomena we need to understand identities formed based on opinions about iconic issues or events (Bliuc et al., 2007; Thomas et al., 2015). Moreover, we need to understand what makes one person more likely to experience group formation, and how these individual differences are shaped by events in the social world (Duncan, 2012).

In terms of differentiating those who identified as supporters (vs. those who do not), being low in legit-imizing ideologies (SDO and RWA) was associated with increased likelihood of self-categorizing as a sup-porter. People who hold entrenched worldviews that support hierarchies (SDO) or see the movement of people as a threat (RWA) are less likely to be a part of the mobilization potential (path a, Figure 1; Klander-mans, 1997; Stewart et al., 2016). Moreover, exposure to powerful images such as that of Aylan Kurdi, appears to undermine group formation for those high in SDO (Figure 2; cf. Asbrock et al., 2012). Thus, although social media plays an important role in pre-dicting group consciousness amongst those who are self-identified supporters, it seems to polarize support and opposition on both sides of the debate.

It has long been observed that specific iconic events (such as those initiated by Rosa Parks, the Greensboro Four or Mohamed Bouazizi) form signal moments,

which catalyze social movements (Castells, 2012; Minkoff, 1997). Such events represent watershed moments in social movement formation, and social media provide ways to rapidly disseminate these events, providing spaces for people to validate their views across geographical, cultural and national boundaries (Smith, Thomas, & McGarty, 2015). Con-trary to accounts that emphasize the superficial nature of engagement online, the popular sympathy gener-ated by the image of Aylan Kurdi provoked change in many countries, confirming the power of social media to shape social change at a global level (path b, Fig-ure 1; Castells, 2012; cf. Morozov, 2011; Smith et al., 2018).

How Social Movements Mobilize Action Within and Across Nations

Our theories and methods need to consider both the factors that promote unity across transnational diver-sity and also those factors that imbue group member-ship with a local (national) flavor. With the exception of outrage (which did not play a uniformly important role), we found that the hypothesized markers of group consciousness were valid for each of the national samples (configural invariance). Social

identi-fication– psychological membership of a group of

peo-ple who support refugees – was also shown to be a

universal (metric invariant) component of the emer-gent group consciousness supporting the core tenet of the social identity approach that social identification promotes perception of similarity and unity in the social movement (Turner, 1991). However, cognitive measures of social identification do not allow for a consideration of the ways in which the experience of group membership may differ, depending on culture, language and history as well as current socio-political realities. Whilst the social psychological literature on collective action scholarship recognizes the importance of social identification, it has engaged rather less with the phenomenology of group membership (but see, e.g., Drury et al., 2005; Duncan, 2012; also Thomas et al., 2009a). It was here that beliefs about the nation’s ability to act effectively (as a coordinated group), as well as emotions that indicate a critical anal-ysis of the (national) group’s position (guilt), more meaningfully provide information about cross-cultural variation in the group consciousness in the different samples (path d, Figure 1).

The results suggested that feelings of guilt and effi-cacy beliefs played a relatively stronger/weaker role in explaining the group consciousness across national contexts. Whilst it is tempting to provide a post-hoc explanation for these differences in terms of the pre-vailing social and geo-political contexts, it is important to note that there can be many different reasons for measurement invariance (see Byrne & Watkins, 2003, for a review). Differences can reflect “real” (genuine) cross-cultural differences and/or be due to measure-ment error or differences in methodology (sampling,

question translation) and it is not straightforward to disentangle these effects (see Byrne & Watkins, 2003; van Zomeren & Louis, 2017). In this case, guilt loaded onto the group consciousness factor to a relatively greater degree in Germany and the UK, suggesting the possibility of different historical (Germany) and/or current political (UK) realities (Vis & Goriunova, 2015). Similarly, the importance of instrumental cop-ing (efficacy beliefs; van Zomeren et al., 2004) was stronger or weaker, possibly reflecting the real barriers to action that confronted the supporter in that context. Future research should include sufficient numbers of Level 2 units (national groups) to use multi-level structural equation models to differentiate item bias from “genuine” cross-cultural difference (as suggested by Davidov et al., 2012).

We also considered whether the mobilization pro-cess outlined in Figure 1 was stable (universal) across the six countries, or whether there was cross-national variation. The results suggested that there was varia-tion in the importance of the iconic image in explain-ing group consciousness (path e, Figure 1), but also that the form of the solidarity (path g, Figure 1) dif-fered across context. However, individual differences played a stable role (path f, Figure 1) and the nature of social media exposure was also invariant (path e, Fig-ure 1). Again, post-hoc explanations are not appropri-ate given that we cannot rule out differences in sample and methodological artefacts for these effects. It may be that the social media exposure was stable across cultures because the phenomenon itself is not bounded by national borders.

Nevertheless, we believe that this method of theo-retically and empirically accounting for cross-cultural similarity and difference using tests of measurement and structural invariance holds great promise. Qualita-tive approaches can provide a rich perspecQualita-tive on the lived experience of group membership (e.g., Drury et al., 2005; Stuart, Thomas, Donaghue, & Russell, 2013) but may also be fraught with difficulty in cross-cultural settings where many national samples are involved. We therefore believe that the method of using tests of measurement invariance could form a useful addition to the social psychological toolkit. This method could also be more widely adopted in social psychological research generally to model, for exam-ple, changes in identity content pre- and post- experi-mental intervention (e.g., group interaction; Thomas & McGarty, 2009), or qualitative changes in the mean-ing of group membership as a group politicizes or radi-calizes (e.g., Thomas, McGarty, & Louis, 2014). Limitations and Future Directions

Social dominance orientation and RWA were each measured with a single item and we do not know if findings would differ had we used the full scales. Given the muted findings for RWA in explaining group consciousness, it may also be the case that RWA is more strongly implicated in the emergence of